Activity Descriptor for Video Sequences

1. A method for describing activity in a compressed video sequence, comprising the steps of: measuring intensity attributes of the activity in the compressed video sequence; measuring direction attributes of the activity in the compressed video sequence; measuring spatial attributes of the activity in the compressed video sequence; measuring temporal attributes of the activity in the compressed video sequence; and combining the measured intensity, direction, spatial and temporal attributes of the activity in a digital descriptor of the activity of the compressed video sequence.

2. The method of claim 1 wherein the compressed video sequence includes video components.

3. The method of claim 1 wherein the compressed video sequence includes audio components.

4. The method of claim 1 wherein the compressed video sequence includes a plurality of frames.

5. The method of claim 1 wherein the compressed video sequence includes a plurality of video objects.

6. The method of claim 1 further comprising: measuring motion vectors representing the activity in the compressed video.

7. The method of claim 6 further comprising: measuring magnitudes of the motion vectors; and averaging the measured magnitudes.

8. The method of claim 6 wherein run-lengths of the motion vectors are measured.

9. The method of claim 8 wherein the run-lengths are classified as short, medium or long.

10. The method of claim 9 wherein the compressed video sequence includes a plurality of frames and run-lengths express the number, size, and shape of moving objects in the frames and distributions of the objects across the frames.

11. The method of claim 1 wherein the spatial attributes are expressed as an angle matrix.

12. The method of claim 1 wherein the compressed video sequence include a plurality of frames and each frame includes a plurality of macro-blocks, and further comprising the step of: averaging the measured attributes over the macro-blocks of the compressed video sequence.

13. The method of claim 1 wherein the compressed video sequence includes a plurality of frames and the attributes are measured for each frame, and further comprising the step of: randomly selecting one frame as representing the activity of the compressed video sequence.

14. The method of claim 9 wherein the compressed video sequence includes a plurality of frames, and each frame includes a plurality of macro-blocks, and further comprising the steps of: counting the total number of macro blocks; counting the number of macro-blocks having with a motion vector ( 0 , 0 ); and expressing the intensity attribute as a ratio of the two counts.

15. The method of claim 1 wherein the compressed video sequence is one of a plurality of compressed video sequences and each of the other of the plurality of compressed video sequences has a respective digital descriptor of the activity of the one compressed video sequence, and further comprising the step of: matching the one compressed video sequence with another of the plurality of compressed video sequences based on the digital descriptors.

16. A method for describing an activity represented in a compressed video sequence, comprising: deriving motion information of the activity representing the compressed video sequence; determining an intensity attribute corresponding to an intensity level of the represented activity, a direction attribute corresponding to a direction of the represented activity, a spatial attribute corresponding to a spatial dimension of the represented activity, and a temporal attribute corresponding to a temporal dimension of the represented activity, based on the derived motion information; and describing the activity with a descriptor corresponding to the determined attributes.

17. A method according to claim 16 , further comprising: combining the intensity, the direction, the spatial and the temporal attributes to form the descriptor.

18. The method of claim 16 , wherein the motion information is derived by extracting information from a motion vector included in the compressed video sequence.

19. The method of claim 18 , wherein: the extracted information includes one of a magnitude and an angle of the motion vector; and the motion vector represents one of a predictive frame and a predictive object included in the compressed video sequence.

20. The method of claim 19 , wherein: the compressed video sequence includes a plurality of macro-blocks of pixel data associated with the predictive frame or the predictive object included in the compressed video sequence; each of the plurality of macro-blocks has an associated attribute; and at least one of the determined attributes includes the attribute associated with one of the plurality of macro-blocks which is closest to an average of the attributes associated with the plurality of macro-blocks.

21. The method of claim 19 , wherein: the compressed video sequence includes a plurality of macro-blocks of pixel data associated with the predictive frame or the predictive object included in the compressed video sequence; and the intensity attribute is determined by computing a ratio of the macro-blocks without a motion vector and the total number of macro-blocks associated with the predictive frame or the predictive object.